摘要
将小波方法引入到深圳宝安住宅房产均价的长期趋势的预测中。利用小波多尺度分析的功能,将非平稳时间序列的住宅房产均价分解成主要趋势和细节两大部分。对主要趋势用二次多项式进行拟合预测,对细节部分用余弦逼近和AR(2)模型进行拟合预测,最后把主要趋势的预测值和细节部分的预测值相加,得到原始时间序列的预测值。并用所建立的模型对深圳宝安2006年每个月的住宅房产均价进行了外推预测。由于这种方法能够准确的提取出房产均价的长期趋势用于研究预测,因此用这种方法预测的平均相对误差比直接用二次多项式进行拟合预测的平均相对误差小0．7％。用所建立的模型外推预测出的2006年1、2、3月份的数值与实际数值相比较,平均相对误差为0．033。误差是比较小的,反映了小波预测方法的有效性。
The wavclet method is introduced to forecast the long - term trend of the housing mean price in Bao an area of Shenzhen. By using the function of wavelet multi - scale analysis, the non - stationary time series of the housing mean price can be divided into main trend part and detail part. The main trend part is imitated and forccasted by quadratic equation , and the detail part is used by cosine to approach and AR（2） model to forcast. At last , the forecasting value of the main trend part and the detail part is added to obtain forecasting results of the original time series. Meanwhile, the built model is used to forecast the housing mean price of every month in 2006 in Bao An area of Shenzhen. Because this method can abstract the long - term trend of the housing mean price accurately to study and forcast, the average relative error of his method is 0. 7% less than the method of directly using the quadratic equation to imitate and forecast. The average relative error is 0. 033 by comparing the foreast value with the real value on January, February and March in 2006. The small error reflects that the wavelet method to forecast the long - term trend of the housing mean price is available.
出处
《陕西理工学院学报（自然科学版）》
2006年第3期25-28,共4页
Journal of Shananxi University of Technology:Natural Science Edition
关键词
小波分析
时间序列
预测
wavelet analysis
time serles
forecasting
